ISSN 0281-9864 The Controlling Function of the Agent in the Analysis of Question-Response Relationships Inger Bierschenk Lund University Sweden KOGNITIONSVETENSKAPLIG FORSKNING Cognitive Science Research The Controlling Function of the Agent in the Analysis of Question-Response Relationships Inger Bierschenk 1987 No. 19 Communications should be sent to: Bernhardt Bierschenk Department of Psychology Paradisgatan 5 Lund University S-223 50 Lund, Sweden Abstract In contrast to traditional linguistic analysis, a model based on the empirical Agent is put forward and tested. A text is regar- ded as an intentionally produced cognitive process. The analysis has to take the Agent (perspective) into account in order for an adequate processing of its Objectives (viewpoints) to come about. Moreover, the model is surface-oriented and assumes the cognitive relevance of an utterance to be defined by the dynamics of the text production and not by artificial semantic criteria. The model has been tested on responses to questionnaire items constructed by a multinational industry. The responses were produced by 35 randomly selected subjects from England, Italy, Sweden, West Germany, and the United States. The differences were tested for power of normal curve test of P.. = P- via arcsine transformation at a = .05. By Agent control of the question-response relationship, differences in coherence could be demonstrated. English and Swedish workers showed a significantly higher coherence with the perspective of the industry, whereas Italian, German, and US workers take up their own Agent function in their response behaviour. The following presentation aims at discussing a novel ap- proach to language analysis. A frame of reference will first be given, which connects to traditional linguistic discussions in the sense that the examples are fabricated. In the description of the qualitative difference between the traditional analysis and the one proposed, the examples are taken from natural text, that is, verbal utterances produced in a natural situation. The work- ing material is Swedish, to which a literal translation into Eng- lish is given. The Sub.ject in the Linguistic Context A central notion in linguistic analysis of language is sub- ject . The subject may be of three kinds, logical, grammatical, and psychological. (The grammatical category formal subject will not be referred to here.) The logical subject is defined according to semantic criteria, e.g., the one who acts in the clause. The anal- ysis of grammatical subject is based on syntactic-morphological criteria and may be identified as that about which something is said. For example, adjectives and some verb forms are inflected in congruence with their grammatical subject. The psychological sub- ject is bound to the initial position of the clause and is said to stand for the psychological attention. The variability of the notion subject is illustrated by the following three Swedish examples: Forskarna diskuterade medvetandet (1) (The researchers discussed consciousness) logical grammatical psychological Medvetandet diskuterades av forskarna (2) (Consciousness was discussed by the researhers) grammatical logical psychological Medvetandet diskuterade forskarna (3) (Consciousness discussed the researchers) psychological logical grammatical Example (1) is a kernel sentence, whose main components are realized in exact correspondence with the model, SVO. This means that the interpretation, the deep structure, coincides with the syntactic organization, the surface structure. Example (2) is a passive transformation of ( 1 ) . The task of a transformation is to reform a deep structure to a surface structure without changing the deep structure, which carries the cognitive meaning of the sen- tence. This is marked in (2) by the logical subject. Thus the deep structure of ( 1 ) and (2), and the surface structure of (1) are iden- tical, while the surface structure of (2) is different. Example (3) is the result of a transformation in which the object has been to- picalized and the subject has been degraded. The initial element now is only a psychological subject, while the grammatical subject has followed the logical according to criteria that are not trans- parent. There is no syntactic-morphological criterion that motivates the analysis. The inflected form of the verb is the same after the topicalization. In the absence of manifest criteria for identifying the " true " subject in sentences like (3) some other basis of interpretation than the cognitive has to be considered, that is, presuppositions. This means that the linguistic context plays a part for interpreta- tion, as does also the linguistic intuition. If the sentence is to be analyzed as isolated from previous ones, that is, presuppositions are inadequate means, then the analysis of (3) should be conceived such that the logical and grammatical subjects both are analyzed according to semantic criteria. A pure syntactic analysis does not seem possible. Obviously it has to be denied that consciousness may be a subject in a semantic or syntactic sense. The primary basis for this conception is our common sense of the " objective world " , which tells that consciousness is something abstract, which cannot be acting and thus not perform anything that could be called dis- cussing. Any objections from the poet or the introspecting novelist would not be acceptable, because everybody knows that those text producers deal with fiction. The conception of consciousness as a true subject is simply counterintuitive. One foundation for this semantic position is case theory (Fill- more, 1968). This theory defines which cases can be realized for types of verbs represented in case- or argument frames, which are part of the lexicon. The bearing argument for case specification is the logical distinction between Animate and Inanimate. Thus the frame for the active transitive verb discuss specifies the sub- ject to Animate and its case is called Agentive or Agent. It is likely that consciousness would be classified as an abstract In- animate, which implies that its case name must be looked for on the next higher level, probably Objective or Object. Several case theories exist, underlining the fact that any unambiguous knowl- edge of the world cannot be referred to. It is also clear that the analysis tends to be obstructed in relation to the abstractness of the language level. The difficulties lie in the requirement of in- terpretation, that the deep structure sense must be tested against the surface structure. It would totally change the situation if the linguist could control the producer of the utterances and ana- lyse his/her subjectivity with objective instruments. Existing me- thods are based on context free sentences in order to function op- timally for objective analyses, which has as its consequence that the linguist's own subjectivity is slipped into the analysis. This methodology is also applied on texts but is no easy way depending on the amount of combinations to represent. Intuition is a central concept in theoretical linguistics. One could say that the collective meaning hierarchies which seman- tics provides for the language elements constitute the stereotyped intuition, that is, a frozen convention. The task of linguistics sofar has been to establish rules for the way in which the conven- tion is used in language perception and production. Only when the language connotations are followed, the meanings of utterances will be correctly interpreted by the language community. Thus language as convention is a central label, too. But at the same time hardly anyone today would argue that human language is something else than a natural phenomenon, which grows with every person's development. The concept of natural language and research related to it has attracted special attention because of the existence of formal lan- guages, computer languages. Within artificial intelligence in par- ticular, it seems important to stress that natural language (as opposed to artificial) is simulated. The entire computer line in language research has come very close to an invalidation of the difference between living and non living systems (B. Bierschenk, 1986; Bierschenk 8c Bierschenk, 1986 a). AI researchers (winograd, 1983) have caused many linguists to act as if they believed that an elegant procedural solution of a sentence analysis, a so called parsing, is the explicit explanation of the human interpretation process (cf. Dresher & Hornstein, 1976). It now seems as if the procedural sequencing of formal analysis has become convention in the description and explanation of the way in which people pick up verbal information. Sofar no serious discussion; goes on about the counterintuitive in equalizing naturalness with conventionality (cf. Pereira & Warren, 1980), as if a natural behaviour could not exist under the conventional surface. We rather should ask ourselves if naturalness is not discovered and lifted up the best way when the convention is known. The linguistic model to be presented here constitutes a refor- mulation of this question into an assumption. It is a model in which consciousness is the true governor. The primary prerequisites of the model are the following: 1. Language and text are expressions of the intention (Agent) of somebody acting, and cannot be isolated from this point of re- ference. 2. The surface structure reflects the cognitive relevance directly. 3. The cognitive relevance is defined with reference to the Agent, and not to any objective world criteria. Expressions of the Agent in Natural Language In the continuing text, examples will be given from authentic material in which an official of the Swedish public sector gives his view of his work situation in a natural discourse. The text, which has been analyzed in its entirety (Bierschenk & Bierschenk, 1986 c) begins with the following sentence: Titta pa hur installningen ar idag (4) (Look at how the attitude is today) The formal analysis shows that the sentence contains two verbs, each of which requires a subject. The attitude is the subject of the clause beginning with how , whereas any subject of look has not been expressed. The conventional analysis method assumes that an imperative sentence like this is the result of a transformation from a statement like You shall look . This means that we have to in- vent a logical subject, which may be you in the natural situation. The person referred to by you is assumed to be the person who is being requested to act by the verb and can therefore be specified as Agent. The object of the looking is the clause that follows. But how to interpret it? The verb i_s does not denote anything active, so the attitude cannot have the same status as you . It is not an Ani- mate either, rather an Objective. The verb is, is called a copula, connecting a nominal with its attribute. The interpretation of the clause should be that it is identical with the attitude today . By that the clause is no clause. A problem that remains to be solved concerning the correct analysis of the real situation is that the factual reference of you must be unambiguously identifyable. The interviewed person makes the following utterance further on in the discourse: Man maste alltsa, om man tar det politiska jobbet, fa (One must thus , if one takes the political job , get) jobba med fragor t ex i ditt bostadsomrade, som (to work with matters e.g. in your living area , which) jag overhuvudtaget kanner att jag berors av. Pa det (I on the whole feel that I am effected by. In this) sattet kan du skapa ett engagemang som du sen kan (way can you create a commitment which you then can) fora vidare till storre fragor. (carry further to bigger matters.) It seems to be an unsurmountable task to specify when the factual referents of the personal pronouns coincide and when not. Instead of assuming that the speaker intends to refer to particular individuals, it is more fruitful to assume that he mixes the pronouns because any individually different Agents do not exist in his mind. He takes the others' perspective and presumes that they would do the same regarding this theme of discourse. Several X-, you- , and one- variables take the same standpoint. Thus only one I_-referent de- fines the text. With this view we do not need to presuppose that the Agent of look is some you with a factual referent but that, instead, any you might take the same standpoint as the Agent, as for example the reader who was not present at the moment of produc- tion. The Agent is determining the perspective. By taking the Agent function into consideration in the analysis of the expression, we are not dependent on a deep structure testing. And, after all, that is the way the text functions in perception. That which acts is the intentionality of the Agent. It is very likely that the use of verbs may be seen as a simi- lar textual phenomenon. It is logically impossible that the verb look can be followed by something as abstract as it really does. But despite that the verb is unconventionally chosen, it functions adequately, that is, as a bridge between the Agent and that which he wants us to focus upon, his viewpoints. So why would we need to presuppose that is in a natural context would have a strictly logic- al meaning? It is more natural to regard it as a bridge by means of which the Agent lets the attitude act towards today . The Agent is the subordinated factor through which the viewpoints of the text are described and related. Each time the Agent uses a verb it is an expression of an intent to shift the perspective. The shift is mani- fested in what will be termed textual agents. The attitude is the first textual agent through which the Agent comes into view. The textual agents are explicitly present in the text in such a way that every clause has an explicit textual agent or a copied one. The Agent is implicitly present and is introduced into the text by certain particular formulations. The convention lying in the syntactical or morphological variation is the key to the presence of the Agent. But not any variation. Four prototypical formulations reveal the Agent, namely (1) the imperative, which has been dis- cussed, (2) the direct question or similar formulations operating by indicators termed clause openers, (3) the passive, and (4) the topicalized formulation with preposition as operator. The distin- guishing of the Agent (perspective) and the Objective (viewpoints) (I. Bierschenk, 1984 a) may be exemplified as follows (X) Titta pa hur installningen ar idag ((X) Look at how the attitude is today) A a 0( A a (x) ska jag (x) hjalpa kommunen (X) shall I (X) help the municipality) Varfor (Why Sen (Then) > Folk tas in som gisslan har nagon gang (People are taken (X) in as hostage here some time) a A nar det behovs (x) (when it is needed (x) a A Pa det sattet (X) kan du (x) skapa ett engagemang (In this way (x) can you (X) create a commitment) A a A a (5) (6) (7) (8) Every verb (a = action) implies a complete A-0 relation (Bier- schenk & Bierschenk, 1976). In (5) the A-0 of the subordination is copied up into the former clause, at the same time as it constitutes an independent relation. The examples (6) and (8) show how the Agent is copied down to keep the perspective constant. In (7) the morphological variant of the verb gives the information that the Agent shall be marked. Concerning the identification and differenti- ation of the O-component, see I. Bierschenk (1984 b). A contrast ive analysis of the imperative (5) was the point of departure for the discussion (4). The following discussion will concentrate on the other three. In terms of transformational grammar, the finite and infinite verb have the same subject and therefore the surface structure is regarded as a transformation. This type is called EQUI-NP deletion. A grammatical analysis of the direct question presupposes that a permutation has operated between the subject and the finite verb and that a question morpheme by means of some transformational stage would be the marker for the place of the deep structure sub- ject. A grammatical analysis has no other possibility than rely- ing on syntactic positions for determining the categories. The syn- tactic-morphological variation is identified by a so called posi- tional scheme, which became one of the descriptive instruments of structural linguistics. However, this scheme should not be under- stood as a Kantian schema but as a frame (B. Bierschenk, 1981; I. Bierschenk, 1984 a), that does not allow functional variation. Table 1 presents the difference between statement and question (6) when arranged into the position scheme valid for the main clause in Nordic languages (Diderichsen, 1962). The positions are arranged into fields, underlining the phy- sical orientation of the model. The positions showing variation are the fundament and the nexus subject. The fundament may be a subject or an adverbial, which, again, demonstrates that some kind of seman- tic specification is presupposed. In the case of a Yes-No question the fundament is empty, while by Wh-question it is filled with the Wh-element (clause opener in the present model). The statement shows that the subject is to be found in initial position. Teleman (1962, p 46) interprets Diderichsen 1 s description of the fundament as a placeholder where certain constituents may be moved " from their normal position " . The normal word order characteristic of the state- Table^. Positional analysis of statement and question according to Swedish word order Fundament Nexus Field Content Field s" 3 v s a V S A I shall (not) help Y Why shall I help Y Shall I help Y (again) Note , v = finite verb, s = nexus subject, a = nexus adverbial V = infinite verb, S = object, A = adverbial complement 10 ment with the subject in the fundament would then be the result of a transformation from the deep structure. However, Diderichsen does not talk about transformations, and it is incorrect to discuss the realizations of positions in process terms, because it blurs the model. The advantage of the position scheme in linguistic descrip- tion is due to its character of control instrument. The observation that certain constituents sometimes shift their position may corre- late with various extralinguistic factors but cannot be explained by them. The function of the fundament as a cue to a correct placement of different clause constituents according to the prescribed order in the scheme is indisputable to anyone who has tried. Its simila- rity with the psychological subject ( 1 , 2, 3) is obvious. Because of its forming function it lies beyond a theoretical description, since any semantic description of the initial position seems im- possible. For instance, what similarity is there between Why and I in Table 1? A position scheme is static. The physical basis of the model may be expressed as (I. Bierschenk, 1984 a) S ^ V -*- (9) The object determines the statement, that is the predicate. Any responsible subject is not supposed to have influence on the state- ment but only to be associatively linked (nexus) to it. The model reflects the fact that, historically seen, the wills of the gods (omens) have been the only possibility of humans to state something and, consequently, to understand something about their world (B. Bierschenk, 1986). with consciousness of the personal responsibi- lity the subject could be expressed in language, which distinguished the environment from the consciousness of it. In physically based models of science, any subject is not allowed to enter into the de- scription of the environment. This view is convention in linguistic analysis, too. There exists independence between subject and predi- cate. All verbal expressions are analyzed as predicates, that is, as if no responsible " expressor " existed. The AaO model the way 11 it is described in B. Bierschenk (1984) and Bierschenk & Bier- schenk, (1986 b) brings out the responsible Agent through the func- tional relationship between the components: A — a — -0 (10) A functional analysis prerequires a steering mechanism to represent the dynamic and process-oriented aspects of natural language. This mechanism gets its expression by the verb. The AaO model is a struc- tural schema in the true sense. This means that even though the components are not always manifested, they are discoverable through the verb, since they are always present. Example (7), the passive formulation, gives the information that the schema is realized in the OaA order by the passive marking morpheme of the verb. The passive is generally conceived of as one of the most basic transformations in which the deep structure object is moved to the position of the subject. The object position is filled by the sub- ject, whose semantic meaning of Agent is marked with the preposition by . Agent deletion may afterwards be applied, by which the new ex- pression implies an Agent in final position. It shall be recognized in spite of unnormal position. Thus Agent marking is morphological in a double sense, while the syntactic criterion is subordinated. When applied on (7) the following change in formulation might be implied: Kommunen tar in folk som gisslan har (11) (The municipality takes in people as hostage here) nagon gang nar kommunen behover det ( some time when the municipality needs it ) Folk tas in som gisslan har (av kommunen) (12) (People are taken in as hostage here (by the muni-) (cipality) ) nagon gang nar det behovs (av kommunen) (some time when it is needed (by the municipality)) 12 This example (12) is atypical in connection with traditional exemplification. Usually examples are of the type John kissed Mary — »- Mary was kissed by John . The natural text example indicates that an Agent cannot be directly generated through the by-insertion after a passive verb. In both clauses there are formulations which make the Agent phrase unexpected or almost ungrammatical. The lo- calizing adverb here may be said to be used with the purpose of specifying the Agent to the organization in whose offices we find ourselves at the moment . The general pronoun it. (formal subject) gives the impression that the organization in an abstract sense is referred to and not any distinguishable person. It is a bit diffi- cult to even add an Agent to this formulation. In recommendations concerning written composition in school, it is often said: Write in the active mood. Passive formulations are difficult to be successful with, since they easily lead to wrong presuppositions of the Agent. On the university level the advice is the opposite: Write in the passive mood. A scientific conduct requi- res distance between text and Agent. The passive is used in all con- texts in which the Agent should not be made conscious for the lis- tener/reader. The obligatory insertion of the Agent that is made by perspective analysis (7) does not imply a complementary addition of a surface structure. It is not a question of filling a position. The (X) is a controller of the point of reference of what is being said so that the viewpoints can be discovered. The placement of (x) therefore is unimportant. It deserves to be pointed out once again the relationship bet- ween the lexical meaning of the verb and its textual function. The verb shall not necessarily be associated with the Agent, despite that the passive marking morpheme ties the two components to each other. The convention is a supporting factor here but does not in- vite to a literal interpretation. As in the case of pronouns, (x) can in one and the same text refer to more than a certain particu- lar source. But here, too, it is fruitful for the analysis to re- gard the variable (X) as the component which controls that the per- spective is invariant. The information that people are taken in as hostage (7) is mediated as the Agent's own perspective, but also as 13 the perspective of the municipality which he represents. The sta- tistic analysis of the perspective of the interviewed person has demonstrated, however, that the Agent's and his organization's perspectives are incoherent. As a consequence of the experience of functional fixation (Bierschenk & Bierschenk, 1986 c, p 14), the Agent's cognitive process ends up in liberation. This person has left the public sector to start his private business. To conceive of the passive formulation as a transformation does not seem natural (Broadbent, 1977). The passive has very few instances in the present text material, which otherwise is richly varied. Some grammatical marking of the Agent with by_ is totally absent, so an example construction does not make any sense. Example (12) gives a hint that the passive should be regarded as complete in its surface structure formulation and that only one morphologic- al criterion is applicable to identify the Agent. In a construction with a passive verb combined with a grammatical Agent the by-phrase should be conceived as a subcomponent of the Objective, termed Ground (I. Bierschenk, 1984 b) and denoting a point of orientation. The prototypical function of the preposition is to be generally orienting. The formulation in example (8) is the result of a topicaliza- tional transformation (3), to speak in traditional terms. Thus to- picalized formulation denotes the surface phenomenon only. The term is connected with the topic-comment distinction in discourse anal- ysis. Topic has a fundament status syntactically and marks the given or known information in the discourse while comment denotes the new or unknown information. An even more marked distinction is that the determination of topic can only be made context dependent ly, that is, in the referential development of the discourse. To identify a topic in a kernel sentence (1) is uninteresting if not impossible. The initial prepositional phrase has a topic- like function and is fairly similar to the psychological subject. But to be identified, it needs not be regarded as an adverbial phrase topicalized from the deep structure, and no morphologically marked reference to known information from preceding clauses is needed either. The preposition is the direct cue to the presence 14 of the Agent. It contextualizes the Agent, which means that the Agent emerges against a background of some empirical experience or lets the textual context form the frame of reference for the con- tinuing development of the process. The process develops after the verb where the function of the prepositions is to differentiate between viewpoints and decide their cognitive function. Experience builds on this differentiation pro- cess. But all experience does not become integrated so that the Agent can formulate it conceptually. The contextualization (8) is an expression of this kind of non-integrated experience. In the anal- ysis of texts of considerable length, the differentiation of the Agent component brings out more information. The imperative and the passive formulations thus express the unconcealed, undifferentiated Agent who encourages the listener/reader to make experiences. The direct clause opener formulation is an expression of Experience (6), since without experience one cannot whether ask questions nor re- late observations in time. The prepositional topicalization is named Context. Both components are subcomponents of the Agent. Expression of Integration There exists a systematic difference between the way in which integrated and non-integrated experience get their verbal expres- sion. Where the Agent is present in the formulation, integration is not expressed. If instead a textual agent is formulated, this im- plies a higher degree of abstraction, which means that an integra- tion may be formulated. The abstraction is the prerequisite of the formulation of the conceptual (cognitive) function of the Agent. In the text, the following example can be found: Jobbarna i tva grannforetag skiter (13) (The workers in two neighbour companies don't give) i varandra (a damn in each other) What has been verbalized between the textual agent and the verb is a postpositional attribute, grammatically explained as a reduced predicative in a relative clause, that is, who are to be found in 15 two neighbour companies . The relative clause formation implies a formulation in which the subject in a subordinate predicative is deleted when it is identical with the subject of the main clause. The copula, which should be as contentless and static as possible, underlines the spatial sense of the preposition: neighbour compa- nies contains the workers . This semantic representation remains after the reduction according to the principle that the deep and surface structures are semantically symmetrical. Transformation is here used in the sense of movement or trans- fer, where the change is of a positional nature. When used in a functional sense, however, transformation denotes a change from differentiation to integration. By that, change is related to de- gree of abstraction, which is asymmetric, and not to a raising of hierarchical level. Moreover, a functional transformation is not derivable. The orientation expressed by the preposition when func- tioning in the differentiation process is changed to the specifica- tion of the textual agent concept. Thus the textual agent should be conceived as a non-divergent whole, that is, a fact, whereas the elements in a process produce information that might give rise to facts. That a text is characterized as meaningful depends on the textual transposition that the elements undergo. In this process, which we may call the cognitive process, experiences and contextual- izing expressions have the function of differentiating it. By con- sidering the Agent's way of differentiating his standpoints, the structure in his viewpoints becomes specified. A rule system has been developed (Bierschenk 8c Bierschenk, 1986 b) and algorithmically tested (Bierschenk & Bierschenk, 1986 c). Ten rules out of a total of fifty identify and differentiate the Agent component, that is, five process rules and five supplementa- tion rules. Experiment This study considers the textual relationship between a ques- tion item and its unrestricted responses in a questionnaire investi- gation. This relationship differs markedly from a dialogue situa- 16 tion in that an interaction cannot take place. The responding per- son has no possibility to share the intention and takes very easi- ly a passive role towards the questioning person. Of course this effect depends to a high degree on the methodology as such, which implies that the subject is objectified and treated as independent of some identity. But the reponse behaviour is also conditioned by the way in which the question has been verbally formulated. This factor has to be accounted for when the subjects are given the free- dom to answer with personal formulations but respond curtly and re- actively nevertheless. Many questionnaire constructors are unaware of the fact that by a certain formulation they may steer the sub- ject to formulate itself coherently with the constructor, that is, to give its view from the perspective of the constructor, something that needs not be the purpose of the investigation. With this back- ground the following hypothesis has been formulated: Hypothesis : Response behaviour is an expression of a subject's objectified perspective of itself. Method Subjects . The subjects in this study make up a random sample of 35 mechanics drawn from an investigation of about 3000 persons from Europe (England (e), Italy (I), Sweden (S), West Germany (G)) and the United States (US). The workers are employed by a Swedish multinational industry. The analysis builds on the responses of 7 persons from each country. The numer is based on the circumstance that only 7 Swedish responses were usable. Therefore the same num- ber of persons was randomly sampled from the other four countries. Materials . The responses used are taken from a questionnaire distributed to the 3000 mechanics by which they have answered questions about their work situation. Question items nos. 4, 5, and 6 in the form have so called unrestricted response alternatives and concern the workers' conduct towards information related to their job function. The responses to the three items taken together have been analyzed elsewhere (Bierschenk 8c Bierschenk, 1987). For this study question item no. 5 has been selected. It is the most general one of the three and includes the other two contentwise. Further, the responses are the most varied to this question. It reads as 17 follows in the four languages represented. (The industry is re- sponsible for variations due to translation. ) Question 5 E/US Do you have any suggestion how getting information to you can be improved? I Hai qualche proposta su come migliorare il sistema di tras- missione delle informazioni al personale d'officina? S Har du nagra ideer/forslag pa hur informationen till dig kan forbattras? G Haben Sie einen Vorschlag, vie man die Nachrichteniibermitt- lung zu Ihnen verbessern kann? The ways in which the mechanics take up the questions are exempli- fied with one response item from each country. The responses may be longer than what they appear here. Response examples E Yes, by telephone US All mechanics used to receive bulletins for our looseleaf binders I Di farle piu semplificate (To make them simplier) S Utbildningstimmar pa arbetstid (Instruction hours during work time) G Ich bin mit der Nachrichtenubermittlung zufrieden (I am pleased with the news transfer) Design and procedure . In the following analysis the question- response relation is regarded as textually bound. Responses with- out an A-component are treated as subordinated to the question in the sense that it constitutes a conceptualized O-component. Its A- component is thereby assumed to be identical with the Agent of the question, which is the operational definition of coherence here. Responses formulated with an A-component are regarded as expressions of integrity. To illustrate the method by which the Agent component has been controlled, Table 2 gives the question in its English ver- sion together with the responses quoted above. The use of the func- tional schema shows some systematic variations. The presence of the Agent implies the infinite verb form. The subordination needs not, as in this case, be marked with to(di) . The response Send a few 18 Table 2 . Analysis of question-response relations by- Agent control A a (X) Do you have any suggestions . . . Yes,(x) (a) by telephone All mechanics used to receive bulletins . . . Di (X) farle piu semplif icate (X) (a) Utbildningstimmar pa arbetstid Ich bin mit der Nachrichteniibermittlung veil trained instructors into the field namely gets the same anal- ysis. A left-out verb always implies an Agent, while finite verb covaries with a textual agent. A verbless response is assigned to the O-component regardless of whether it starts with a preposition or not. The principle here is that no textual element can be regar- ded as a contextualization (8) or an integration (13) as long as a verb does not mark a perspective. The response Yes , conditioned by the question type, has the same function as to, that is, to be the start operator of the response. In all the responses in fact a Yes is implicit. Thus as single response to the question, Yes would have been nonsense information. The responses of question no. 5 have been analyzed according to the schema presented in Table 2. Only the first conceptualization of the responses has been of in- terest, since the respondent's way of taking up the question is ob- served initially. Results 34 out of 35 subjects have given a response to question no. 5. The missing response belongs to the English text. For a comparison between the responses, the proportion coherent responses of the number of responses per country was measured. The result is presen- ted in Table 3. As the Table shows, there is a clear difference between England and Sweden on one hand, and Italy, West Germany and the United States on the other. A test of the proportional differ- 19 Table 3 . Textual coherence between question and responses in European and American text Country Proportion England .83 Italy .43 Sweden .86 United States .43 West Germany .43 ence at a = .05 shows that the power of the effect size is/J= .87, which is a considerable contrast (Cohen, 1969, p 183). The result shows that the English and Swedish mechanics objectify themselves when responding, while Italian, West German and US mechanics sub- jectify themselves, that is, they take advantage of their Agent function in their response behaviour. Discussion The purpose of this article has been to highlight the Agent concept as steering component in linguistic analysis of language and text. An Agent-based analysis prerequires that the text produ- cer's intention can be incorporated into the model, which makes the model functionally operating. This has as its consequence that traditional grammar, as also its transformational variant, cannot be used to explain utterances, since they are based on a positional outlook. Further, the Agent model is surface-oriented in the sense that it is the perspective in an utterance that is directly read out of the verbal formulation. The relevance of the representation does not have to be interpreted, which is the traditional way of synthesizing, with the linguistic point of reference in semantic objective world criteria. Through a strict application of the func- tional operations of the model the empirical Agent's perspective on the viewpoints (0-component) developed in the text emerges. 20 Moreover, the analysis distinguishes the dynamic (transpositional) changes taking place in the development of text by making visible in the presentation the Agent's shifting degrees of experience. One such distinction is the direct question, in which the Agent occupies the A-component, compared to the case when an integrated concept takes the Agent's function. When the model is used in the analysis of question-response relations this distinction becomes visible in such a way that cer- tain responses relate to the perspective of the question (coher- ence) while others express an individual (integrated) perspective. The result of the analysis shows that English and Swedish workers to a considerably higher degree than Italian, West German, and US workers respond to the industry's question as if they conceived of themselves as objects in the situation, in that their perspective is coherent with that of the industry. That they do not take advan- tage of their Agent function should be seen as their expression of not taking up any independent standpoint. Their response behaviour points at a more frequently used habit of subordinating themselves to an authority. It is noteworthy that only England and Sweden are monarchies and have strong labour unions. It is moreover a known fact that Sweden is strongly centralized and that the subjects of this society have been used to rely on " the strong society " in matters of great concern. Another study using this material, although for a different purpose, has shown (Bierschenk 8c Bierschenk, 1987) clearly distin- guishable differences in mentality between these five countries. While Swedish workers show a passive conduct towards information from the industry and English workers are critical to the mediation of information, the interest to learn and to reach mastery in their jobs were typical of Italian, West German and US workers, as also their wish to contribute to constructive cooperation with their employer. Thus the analysis has confirmed the hypothesis that the taking up of an object role can be read from the formulation of a quationnaire response. The result illustrates (1) the importance of formulating question items so that the responding person will not be unintentionally steered to the objectification of himself, and 21 (2) the possibility to discover by means of Agent control the psy- chological reality as it is formulated directly in language. References Bierschenk, B. (1981). Conceptions of cognitive functions in a science of knowing. Didakometry (63). Bierschenk, B. (1984). Steering mechanisms for knowability. Kogni - tionsvetenskaplig forskning ( 1 ) . Bierschenk, B. (1986). The cult of understanding. Kognitionsveten- skaplig forskning (15). Bierschenk, B. & Bierschenk, I. (1976). A computer-based content analysis of interview data . Lund: Gleerup. Bierschenk, B. 8c Bierschenk, I. (1986). Concept formulation. Part I. The phenomenon of cognition. Kognitionsvetenskaplig forsk- ning (10). (a) Bierschenk, B. & Bierschenk, I. (1986). Concept formulation. Part II. Measurement of formulation processes. Kognitionsvetenskap- lig forskning (11). (b) Bierschenk, B. & Bierschenk, I. (1986). Concept formulation. Part III. Analysis of mentality. Kognitionsvetenskaplig forskning (12). (c) Bierschenk, B. & Bierschenk, I. (1987). Consciousness as a function of knowledge and culture. Kognitionsvetenskaplig forskning (18). Bierschenk, I. (1984). The schematism of natural language. Kogni - tionsvetenskaplig forskning (2). (a) Bierschenk, I. (1984). Intended predication. Kognitionsvetenskap- lig forskning (5). (b) Broadbent, D.E. (1977). Levels, hierarchies and the locus of con- trol. Quarterly Journal of Experimental Psychology , 29 , 181-201. Cohen, J. (1969). Statistical power analysis for the behavioral sciences . New York: Academic Press. Diderichsen, P. (1962). Elementaer dansk grammatik (3rd ed.). Kjzfoenhavn : Gyldendahl . Dresher, B.E. & Horns tein, N. (1976). On some supposed contribu- tions of artificial intelligence to the scientific study of Ian- 22 guage. Cognition , 4, 321-398. Fillmore, C. (1968). The case for case. In: E. Bach & R.T. Harms (Eds.), Universals in linguistic theory (pp. 1-88) . New York: Holt. Pereira, F.C. & Warren, D.H. (1980). Definite clause grammars for language analysis. A survey of the formalism and a comparison with augmented transition networks, Artificial Intelligence , 13 , 231-278. Teleman, U. (1972). 0m Paul Diderichsens syntaktiska modell. In: U. Teleman, Tre uppsatser om grammatik (pp. 33-57 ). Lund: Stu- dent lit t era tur. Winograd, T. (1983). Language as a cognitive process. Vol.1: Syn - tax . Reading, MA: Addison-Wesley.